An enhanced approach to predict permeability in reservoir sandstones using artificial neural networks (ANN)
- PDF / 5,514,203 Bytes
- 15 Pages / 595.276 x 790.866 pts Page_size
- 71 Downloads / 263 Views
ORIGINAL PAPER
An enhanced approach to predict permeability in reservoir sandstones using artificial neural networks (ANN) Joel Ben-Awuah 1 & Eswaran Padmanabhan 2
Received: 26 June 2015 / Accepted: 17 March 2017 # Saudi Society for Geosciences 2017
Abstract A key challenge in the oil and gas industry is the ability to predict key petrophysical properties such as porosity and permeability. The predictability of such properties is often complicated by the complex nature of geologic materials. This study is aimed at developing models that can estimate permeability in different reservoir sandstone facies types. This has been achieved by integrating geological characterization, regression models and artificial neural network models with porosity as the input data and permeability as the output. The models have been developed, validated and tested using samples from three wells and their predictive accuracy tested by using them to predict the permeability in a fourth well which was excluded from the model development. The results indicate that developing the models on a facies basis provides a better predictive capability and simpler models compared to developing a single model for all the facies combined. The model for the combined facies predicted permeability with a correlation coefficient of 0.41 which is significantly lower than the correlation coefficient of 0.97, 0.93, 0.99, 0.96, 0.96 and 0.85 for the massive coarse-grained sandstones, massive finegrained sandstones-moderately sorted, massive finegrained sandstones-poorly sorted, massive very fine-
* Joel Ben-Awuah [email protected]
1
Department of Chemical and Petroleum Engineering, UCSI University, Taman Taynton View, 56000, Cheras Kuala Lumpur, Malaysia
2
Department of Geosciences, Faculty of Geosciences and Petroleum Engineering, Universiti Teknologi PETRONAS, Bander Seri Iskandar, 32610 Seri Iskandar, Perak, Malaysia
grained sandstones, parallel-laminated sandstones and bioturbated sandstones, respectively. The models proposed in this paper can predict permeability at up to 99% accuracy. The lower correlation coefficient of the bioturbated sandstone facies compared to other facies is attributed to the complex and variable nature of bioturbation activities which controls the petrophysical properties of highly bioturbated rocks. Keywords Artificial neural network modelling . Reservoir sandstone facies . Bioturbation . Porosity . Permeability estimation . Reservoir rock quality
Introduction Reservoir characterization includes a detailed evaluation of reservoir facies and the corresponding petrophysical properties. A key challenge in the oil and gas industry is to be able to predict key petrophysical properties such as porosity and permeability in the subsurface (Goncalves et al. 1997; Khidir and Catuneanu 2010). Petroleum reservoirs are generally heterogeneous and complex with fundamental properties such as porosity and permeability typically distributed in a non-uniform and non-linear manner. Heterogeneity in facies distribution and facies charact
Data Loading...
